04/11/2020 | Sébastien Gagelin, Expert en data intelligence, Hardis Group

Digital twins might sound like something out of science fiction, but they’re already being used in a wide range of applications. The technology can help to eliminate costly and frustrating gray areas—provided it’s applied to suitable projects and backed by the right software and infrastructure.

What is a digital twin?

In simple terms, a digital twin is an exact, faithful digital representation of a product, process, or organization. Achieving this degree of precision, at a given point in time and as reality evolves, means capturing real-world data. Ideally, this data should come from sensors or other connected objects. Where that’s not possible, fixed or on-board cameras are used instead. 

Once the digital twin has been defined with real-world data, it can be used to analyze, to diagnose, and to develop models. The technology is especially useful for forecasting—building predictive scenarios, running simulations by testing different variables, using expected results to inform decision-making, and even triggering automatic corrective action.

The sheer number of use cases for digital twins is staggering: from process analysis and improvement, to automatic detection of anomalies and security breaches, and end-to-end traceability. Let’s look at an example of a process from the world of logistics: the receipt of raw materials, semi-finished products, and products from suppliers. Incoming pallets often pass through buffer zones, where no digital data is generated (no scanning, hence no data reporting). But with cameras installed in these buffer zones—linked to an object detection model—it becomes possible to generate data and, therefore, to track indicators like buffer zone capacity, storage duration, number of pallets passing through each day, and so on. There’s also the option to set up alerts—for instance, when pallets spend too long held in the buffer zone.

Looking further ahead, we’ll likely see digital-twin projects spanning entire ecosystems (i.e. multiple businesses and organizations) in the race for value chain optimization. In the meantime, individual links in the chain can harness the technology’s potential to drive process improvement according to their own challenges, context and imperatives.

Digital twins: the ins and outs of data sources, software, and architecture

It takes a vast amount of data to set up and operate a digital twin that’s both realistic and useful. This data typically comes from a number of sources—from structured databases, to unstructured and heterogeneous sources like call-center conversation recordings, as well as data from connected objects and camera-based object recognition systems.

Analyzing and processing this data involves AI technologies (machine learning and deep learning), since this is the only way to convert unstructured information into structured data that can be used by business and BI applications, as well as diagnostic, simulation, and other associated systems (which might also be AI-based).

Putting the specifics of requirements and systems aside, one thing’s for certain: the amount of data that needs processing is increasing exponentially. According to the International Data Corporation (IDC), data volumes have grown 50-fold in the past five years alone. As such, getting the underlying system architecture right is critical to the success of a digital twin project. Although Cloud-based resources work for smaller pilot projects, the sheer volume of data involved in full-scale deployment requires a mix of Cloud- and Edge-based architectures. Here, storage and processing happens locally—close to the sensors and connected objects (the Edge part)—and the output data and reports are sent to the Cloud. 

Eliminating costly gray areas

While digital twins have no end of uses, the technology is gaining ground in front-line operational processes (where data is sorely lacking) and, more generally, in areas where emerging technologies like IoT and cognitive services have the potential to solve routine problems. Some examples—again from the world of logistics—include end-to-end traceability of parcels in a warehouse, and anomaly detection (damaged boxes, inconsistent fill rate, incorrectly positioned AGVs, and more).

For businesses, the key challenge lies in priority-setting: selecting projects where digital twins will deliver substantial gains, dramatically reduce costs, or enhance brand image (for instance by reducing errors, raising service standards, or improving communication with customers). 

In other words, the only way we’ll see ever more digital twin projects is if businesses take a pragmatic approach, thinking in terms of their specific needs and objectives. And eventually—perhaps in the not-too-distant future—this development will lead to the advent of collaborative projects.